Visual China
With the rapid integration of global technology and medicine, the application of AI technology in the medical field has triggered profound industry changes. An academic conference jointly organized by the journal Nature and the Eye Hospital of Wenzhou Medical University was recently held in Wenzhou. Zhang Cong, an algorithm scientist at Beijing Eagle Eye Technology, delivered a speech entitled "Medical AI Chips Empower Eye Health and Chronic Disease Management Applications".
At the "2024 11th 301 Health and National Health Action (Xiong'an) Forum and the 13th Academic Forum of the Director of the Physical Examination Center" held at the same time, Professor Zhang Qun, director of the Health Management Center of Jiangsu Provincial People's Hospital and vice chairman of the Health Management Branch of the Chinese Medical Association, as a cooperative doctor of Eagle Pupil, also delivered a speech entitled "AGI Empowers Health Management of Chronic Diseases", which elaborated in depth how AGI provides innovative solutions in chronic disease management.
The two speeches that echoed before and after gave the perspective and position of Eagle Eye Technology in the process of artificial intelligence empowering medical health from the level of technical algorithm change and case study.
How Eagle Eye Technology realizes the integration of diagnosis and treatment through algorithm upgrading
As an algorithm scientist, Zhang Cong mentioned the technological innovation and exploration of Eagle Eye Technology.
"We use a deep learning model based on multi-task classification to describe most of the fundus disease diagnosis tasks in a unified manner, and improve the understanding and expression ability of the large model on fundus images. However, there is an interpretability problem with this kind of classification task, and it is impossible to know exactly why the model prediction is right or wrong.
To this end, Eagle Eye Technology has developed a SAM lesion tissue segmentation model through algorithm upgrades, and carefully segmented the key information in the fundus picture, such as blood vessels, optic cups and plates, and key lesions, such as bleeding, exudation, cotton wool spots, etc., effectively improving the interpretability of the model and realizing the transformation from black box to white box. This work has been published in MICCAI, the leading conference in the field of medical imaging. ”
Through the improvement of deep learning models, Eagle Eye Technology has achieved remarkable results in the diagnosis of 14 common fundus anomalies, with an average AUC of 0.968. The results were also published in the authoritative medical journal The Lancet. Eagle Eye Technology's fundus AI system saves 75% of the screening time, significantly improving the work efficiency of doctors, and empowering primary medical personnel to conduct efficient and accurate fundus screening.
"Although the accuracy of the current detection rate is high, Eagle Eye Technology is not satisfied with the status quo, and considers using the accumulated large number of fundus image cohort data to predict the development of future fundus diseases of patients and achieve early intervention. The company has established a GenAI model to predict future changes in fundus images of high myopia. ”
Eagle Eye Technology uses fundus images to evaluate systemic chronic diseases, pioneers the evaluation of thyroid disease and high-altitude erythrocytosis, and is at the leading level in the fields of hypertension, diabetes, and gestational hypertension. Eagle Pupil not only assesses the current situation, but also works on predicting future disease development, adding multimodal data to assess cardiovascular and dementia risk based on the fundus. In addition, breakthroughs have been made within the company in predicting the development of hypertension and diabetes, and related work has been published.
Eagle Eye Technology's AI, which empowers eye health and chronic disease management services, is also applied to the overall solution of intelligent diagnosis and treatment integrated equipment. Through multiple rounds of system iteration, 1-minute non-invasive detection has been realized, which has greatly improved the detection efficiency and has served more than 30 million people. At the same time, the company's accumulation of a large amount of eye health-related data further enhances the effect of the AI model. In the years of service, the company has found 27,973 major positive cases, saving the lives of many users in a timely manner.
How Retinal AI Makes Invisible Risks "Visible"
The retina is the only tissue in the human body that can directly observe blood vessels and nerves non-invasively, and contains a wealth of biological properties and health information. Artificial intelligence is capable of conducting health assessments through the retina through pixel-level image processing and deep learning.
In recent years, the scientific research strength represented by Eagle Eye Technology has received more and more extensive attention, and retinal AI technology is revolutionizing the risk assessment method of cardiovascular disease, providing a non-invasive and efficient way to predict and manage cardiovascular health.
Using fundus photographs, retinal AI avoids many of the invasive metrics required for traditional assessments, such as blood pressure, blood lipids, and diabetes status. According to the ICVD prediction results published in Science Bulletin, the team of Tongren Hospital and Peking University Clinical Research Institute verified the results, and the 10-year cardiovascular disease risk assessment of the Chinese population using the deep learning algorithm of retinal images showed that the AUC of the model in screening critical/moderate and above (≥5%/≥7.5%) ICVD risk groups was 0.971, respectively and 0.976, which played a positive role in early detection of high-risk groups of cardiovascular disease, guidance of active intervention, monitoring and evaluation of intervention effects.
A nationwide multi-center study further demonstrates the effectiveness of retinal AI technology. From January 2022 to March 2024, the project "Multicenter Study on Diabetic Retinopathy Screening in Physical Examination Populations" was organized by the China Health Promotion Foundation, provided academic support by the Health Management Branch of the Chinese Medical Association, and led by the School of Medicine of the PLA General Hospital.
After conducting extensive scientific research, Professor Zhang Qun has concluded that AI technology can not only identify fundus diseases, but also predict the risk of future chronic diseases, revealing the risk of other chronic diseases such as anemia and dementia. She gave an example of a real-world case that demonstrates the potential of retinal AI in chronic disease health management.
A 45-year-old man with a history of diabetes mellitus and irregular medication is being used. With the application of retinal AI technology, he, like hundreds of patients, began to advance the treatment process and improve treatment awareness. This change reflects the shift from the attitude of "not paying attention and not managing" to "attracting attention and opening up management", and when necessary, reaching the stage of "not ideal, strong management". Specific data showed that in the diabetic retinopathy test in November 2022, the number of hemorrhage points was 5 with a total area of 0.015 mm², and by December 2023, the number of hemorrhage points increased to 15, and the total area was only 0.081 mm². Because of AI retina, there is specific quantitative data, which provides a better basis for diagnosis and treatment of physical examination.
This is not an isolated case, hundreds of patients who have benefited from the benefits of health education and retinal AI monitoring have begun to raise their awareness of treatment, from the attitude of "not paying attention, not managing" to "attracting attention, initiating management", and when necessary, reaching the stage of "unsatisfactory, strong management".
In the development journey of AI, AI is shifting from narrow (ANI), also known as weak AI, to general purpose (AGI) or strong AI. ANI focuses on performing specific tasks, such as language translation or image recognition, while AGI simulates human intelligence and is able to understand, learn, and process a wide range of problems.
From the speeches of Zhang Cong and Zhang Qun, we can clearly see the power of artificial intelligence technology, as well as the changes in world view and values brought about by living cases. AGI technology is gradually changing the management of chronic diseases.
In the future, AI healthcare can provide comprehensive solutions ranging from pre-diagnosis information collection to intuitive display of disease progress. In chronic disease management, AGI can also carry out intelligent pre-consultation, form pre-consultation medical records, strengthen patient education, and improve patient compliance. In addition, AGI has also promoted the progress of smart scientific research. Through automated data processing and analysis, AGI helps researchers greatly improve research efficiency and quality in data collection, statistics, and mapping. The smart scientific research platform also covers functions such as literature retrieval, journal query, and paper duplicate checking, which greatly simplifies the complexity of scientific research work.
Recently, under the guidance of the Beijing Haidian District Health Commission and Zhongguancun Science City, Eagle Eye Technology and 48 primary medical institutions in Haidian District jointly launched the construction of the "Artificial Intelligence Fundus Diagnosis and Treatment Collaboration Platform". This platform is not only the first instance in the country to be deployed privately in the medical alliance region, but its integrated AI technology will greatly enhance the ability of general collaboration.
Relying on the artificial intelligence fundus diagnosis and treatment collaboration platform, Eagle Eye can achieve three major functions for Haidian, provide professional ophthalmologists for the grassroots to visit and teach, and improve the professional ability of general practitioners in the diagnosis of eye-related diseases in primary medical institutions; secondly, provide professional, standardized and long-term training on eye-related diseases to improve the diagnosis ability of fundus diseases in primary medical institutions; and also provide eye disease screening and referral and treatment services for serious fundus diseases for primary medical institutions.
In the new era of artificial intelligence, artificial general intelligence (AGI) is revolutionizing the health management model for chronic diseases. The integration of AGI into the medical system will greatly improve the accuracy and efficiency of diagnosis and treatment if it can be deepened in the standardized tumor screening and fine screening process, and effectively combined with smart wearable devices.